A Multi-Agent Platform to Support Knowledge Based Modelling in Engineering Design

Abstract

Nowadays engineering design process requires the involvement of multiple partners from multiple disciplines throughout the Product Life Cycle (PLC). Consequently, the construction of numerical models became a difficult task due to the distribution of experts. This article proposes an agent based approach to support a coherent know-how elicitation, to enrich design problem analysis, based on the re-use of experiences and their storage in a standardized knowledge base. A set of Tutor-Agents (TAs) aid experts in the knowledge modelling process focusing on Variables, Domains and Constraints as a key component of engineering knowledge. A shared and coherent knowledge base is the main purpose of the proposed Multi-Agent System (MAS). The interaction among agents enables to highlight potential incoherencies during the modelling process to avoid inconsistent information. The Multi-Agent approach is implemented in a software prototype and a knowledge base can then be constructed, providing standardized Product Life Cycle (PLC) constraints (based on the product related knowledge) for creating models to be analyzed by traditional inference engines such as Optimization solvers, Constraint Satisfaction programming, etc.